Clarify implicit roles for <th>

[AlexLloyd0]

The implicit semantics for th are described as

If a descendant of a table element, role=columnheader or rowheader or cell (if not a header).

If a descendant of a table element with role=grid or treegrid, role=columnheader or rowheader or gridcell (if not a header).

But it's not clear under which conditions a <th> should take each role.

https://github.com/w3c/html-aria/issues/61#issuecomment-294574702 suggests an algorithm. Chrome behaves differently to this though, something like:

• If the element has a <td> sibling, it has role of rowheader
• Otherwise, it has a role of columnheader

[scottaohara]

HTML AAM / the HTML spec would be the places to call this out. We can point to the HTML spec / HTML AAM for more info, though. But I do think it's worth re-investigating if th is exposed as a cell in any modern browsers.

[AlexLloyd0]

Ah yes, it is already in the HTML AAM spec, in the th entries at https://www.w3.org/TR/html-aam-1.0/#html-element-role-mappings. I'll create a PR adding a link to that spec.

As for whether browsers ever expose th as a cell, I've found one case in which Chrome (incorrectly) calls a <th> a gridcell (below) but haven't found a cell.

<table>
  <tr>
    <td>d</td>
    <th>h</th>
  </tr>
  <tr>
    <td>d</td>
    <td>d</td>
  </tr>
</table>

[scottaohara]

yes, i've noticed chrome having some odd roles exposed in the browser panel lately. haven't gotten around to seeing if they're incorrectly exposed at the platform level or if it's just the a11y panel in the browser that's quirky here. will investigate that later...

anyway, thank you logging this / making the pr.

[scottaohara]

closed via #246